A Gradient-Descent Optimization of CO-CO-NO Emissions over the Paris Megacity─The Case of the First SARS-CoV-2 Lockdown.

Urban greenhouse gas emissions monitoring is essential to assessing the impact of climate mitigation actions. Using atmospheric continuous measurements of air quality and carbon dioxide (CO), we developed a gradient-descent optimization system to estimate emissions of the city of Paris. We evaluated our joint CO-CO-NO optimization over the first SARS-CoV-2 related lockdown period, resulting in a decrease in emissions by 40% for NO and 30% for CO, in agreement with preliminary estimates using bottom-up activity data yet lower than the decrease estimates from Bayesian atmospheric inversions (50%). Before evaluating the model, we first provide an in-depth analysis of three emission data sets. A general agreement in the totals is observed over the region surrounding Paris (known as Île-de-France) since all the data sets are constrained by the reported national and regional totals. However, the data sets show disagreements in their sector distributions as well as in the interspecies ratios. The seasonality also shows disagreements among emission products related to nonindustrial stationary combustion (residential and tertiary combustion). The results presented in this paper show that a multispecies approach has the potential to provide sectoral information to monitor CO emissions over urban areas enabled by the deployment of collocated atmospheric greenhouse gases and air quality monitoring stations.